Every team's agents in one list
Each agent shows the team, the services the agent covers, the owner and the live version.
Build and govern agents for production.
Agents built in Agent Studio start from the production context graph and act within your policies and control.
Pick the tools, skills and hand-offs
Choose the tools and skills each agent uses, and the agent that takes over when no skill applies.
Policy on every production tool call
Each tool call starts from the right production context. Read tools run. For write tools, your policy decides: allowed, needs approval or denied. Every call is recorded.
Run history for every agent
Each run shows the Slack thread that asked and the version that answered.
Each agent is a file
Instructions in Markdown, settings in YAML: tools, approvals, skills and hand-offs. Edit the form or the file, and keep the file in Git.
Every change is a version with a diff
Compare any two versions field by field or as a file diff before anything goes live.
Publish and roll back with checks
Each release checks the version, the tools and the tests, and records who released and why.
An agent card other agents can read
Each agent publishes an A2A card with what the agent answers and example requests.
What running agents in production takes
NeedBuild in-houseWith Agent Studio
ContextEach agent wires up data access on the sideEvery agent reads the production context graph
PolicyChecks in the prompt, where the model can ignore themAllowed, needs approval or denied for each tool, checked outside the agent
ReleasesPrompts edited in placeVersions promoted or rolled back with a reason
AuditLogs spread across toolsEvery run and action with the asker, version, policy result and approver
Cost of ownershipA platform team to build and run all of the aboveIncluded in the platform
Frequently asked questions
Which plans include Agent Studio?
Both plans. Pay as you go and Enterprise include Agent Studio, and your own agents through skills, MCP or the API.
Can we bring agents built elsewhere?
Yes. Agents built with the Claude Agent SDK or imported from LangChain sit in the same directory and run under the same policy as NOFire AI agents.
What can an agent change?
Each write tool starts at needs approval, and a person approves in Slack. Your policy can allow a tool for a service and environment.
Can we manage agents with GitOps?
Yes. Each agent is a versioned definition. Keep the definitions in Git, review changes in pull requests and publish from your pipeline through the API.
Can one agent ask another?
Yes. An agent asks the agent of the team that owns a service, and the answer comes back under that team's policy. Each agent lists the agents allowed to ask.